Mikhail Matovnikov Central Economics and Mathematics Institute, Russian Academy of Science
BANKING SYSTEM OPERATIONS DURING TRANSITION TO MONETARY STABILITY
Final Report
Moscow 2001
CONTENTS
1. Abstract... 3
2. Introduction ... 4
3. Review of literature / Conceptual framework ... 5
4. Theoretical model ... 9
Influence of Inflation rate change ... 10
Influence of real exchange rate change ... 11
Influence of credit growth... 12
5. Data description and choice of explanatory variables ... 13
6. Model specification ... 23
7. Estimation results ... 24
8. Conclusions ...Ошибка! Закладка не определена. 9. Bibliography ... 24
Cross-country research (listed by year of publication) ... 29
Studies of individual bank failures (listed by year of publication) ... 30
Data on banking crisis episodes ... 30
Other studies... 31
Appendix 1. Estimation results for cross-country sampleОшибка! Закладка не определена. Appendix 2. Banking crises in 1970-1999 ... 35
1. Abstract
The model presented in the paper fills the gap between theoretical discussions and empirical evidence on the macroeconomic reasons of the banking crises.
It is argued that opposite dynamics of some indicators for different crisis syndromes are responsible for the fact that most econometric studies report insignificance of some indicators known from case studies to be important. In fact influence of these factors is not "captured by other indicators" as it is often suggested but "cancelled out" when all crises types are treated as essentially the same.
Sample of countries was selected by pooling countries with more then 2 consecutive years with inflation rate above 40% p.a. The sample includes 46 countries.
Of these countries only 2 - Suriname and Sudan did not have banking crisis (most likely due to data omissions).
Economic discussion suggests that different crisis types are characterized by different set of pre-crisis macroeconomic indicators development. We suggested that along with traditional measures of banking crisis risk factors some factors work in combination with other factors. Namely we suggested that real exchange rate change appreciation is dangerous if accompanied with fast credit growth. And in situation of modest credit growth more danger comes from real exchange rate depreciation. It is argued that it is this opposite influence is responsible for failure of other studies to find support for well known from case studies importance of real exchange rate dynamics in run-up to the crisis.
The other two-sided factor is credit growth. In conditions of high inflation it supports income base of banks and diminishes probability of the crisis. At the same time credit growth during low inflation in some conditions (namely real exchange rate appreciation - see above) may well end with a banking crisis.
Based on this theoretical discussion we split some variables to capture their opposite influence on provability of banking crisis in different syndromes and constructed variables specific for every crisis type.
Logit regression model showed significant expected different signs before variables constructed to model different crisis scenarios. Results support importance of different indicators in different crisis syndromes.
2. Introduction
Studies of the banking crises proliferated in the recent decade with high incidence of banking crises in developed and developing countries. The literature on banking crises is clearly split into two camps. One camp based on macroeconomic cross-country data explores uniform reasons of the crises, the other based on bank-specific and country-specific data presents case studies of banking crises.
One of the striking results is failure of cross-country studies to show significance of some of indicators reported to be very important is case studies. The most notorious example is insignificance of devaluation in cross-country regressions.
This study attempts to overcome this problem by introducing a new model. In this model it is presumed that banking systems are susceptible not to isolated external shocks but rather to complex crises syndromes.
Each syndrome is characterized with common dynamics of macroeconomic indicators in the run-up to the crisis. But some syndromes may involve opposite dynamics of some indicators which in both cases may be significant.
Understanding of different nature of banking crises during and after disinflation is very important for timely preparation for corrective action in implementation and design of disinflation programs. One of important inferences from the presented model is that transition from high inflation tends to be accompanied not be one as it is generally
accepted but by three different types of banking crises that may occur one by one or some crises may be skipped.
This paper is structured as follows. Section 3 presents review of literature and conceptual framework of the study. 4th section presents then data and variable selection.
5th section presents econometric model, and finally estimation results are presented in section 6. And section 7 concludes the paper.
3. Review of literature / Conceptual framework
Empirical studies of the causes of the banking sector problems are divided into two camps. The first (the older one) typically uses data for individual banking crisis and analyses isolated cases of banking problems. Some studies in this group try to generalize inferences from case studies (Lindgren, Garcia and Saal (1996) and Caprio and Klingebiel (1996)).
The second camp explores macroeconomic factors of banking crises and uses cross-country sample data. Due to availability of data on banking crises experience of different countries after Lindgren, Garcia and Saal (1996) and Caprio and Klingebiel (1996) studies, the second approach gained more popularity in recent years.
The shocks associated with episodes of banking sector problems highlighted by the literature include cyclical output downturns, terms of trade deteriorations, devaluations, declines in asset prices such as equity and real estate (Caprio and Klingebiel, 1996, Lindgren et al., 1996, Kaminsky and Reinhart, 1996).
Banking sector problems may also follow successful stabilization in countries with a history of high inflation. As shown by English (1996), chronic high inflation tends to be associated with an overblown financial sector, as financial intermediaries profit from the float onpayments. When inflation is drastically reduced, one of banks' main sources of revenue disappear, and generalized banking problems may follow.
Banking sector difficulties in Brazil and Russia have been explained in this way (Lindgren et al. (1996), Dmitriev et. al. (1996)).
Credit booms as factors for future crisis are emphasized in Gavin and Hausmann (1996), Kaminsky and Reinhart (1996), Honohan (1997).
To summarize, most commonly reported macroeconomic causes of banking crises fall in 3 groups:
(1) sharp fall of inflation rate, (2) unsustainable credit booms,
(3) balance-of-payments problems and sharp changes of assets prices.
Case studies results are only partially supported by econometric cross-country studies. Caprio and Klingebiel (1996) question relationship between banking crises and previous credit booms. No study reported support for importance of devaluation. This relationship is so obvious both from case studies and theory that researchers typically in search for excuses explain such results in the way that influence of devaluation is captured by other factors (see e.g. Demirguc-Kunt and Detragiache (1997)).
Cross country studies usually try to directly estimate the probability of a banking crisis (using limited-dependent variable models) and identify the variables that statistically aid in predicting crises. Empirical methodology of these studies differs a lot, although there is a tendency to use some type of logit (or probit) specification of the model:
∑
==
nj i
crisis
iP
0
exp( ) ) ) exp(
( x
x α α
,where X is a vector of independent crisis factors.
In regressions dependent variable is a dummy that takes value 1 for crisis and 0 otherwise. In multinominal models (Hardy and Pasarbasioglu, 1998, 1999) dummy
takes different values for pre-crisis years, severe problems and systemic crisis, and 0 otherwise.
Inconsistency between results of case studies and empirical econometric studies to out opinion arises because the latter treat all crises as essentially the same phenomena with common set of reasons. In contrast, case studies show that banking crises erupt rather as a result of crisis syndromes where different factors work in combination to produce the crisis.
When presumption of essentially the same nature of all banking crises is discarded, obtained results seem natural. If different crisis syndromes may be characterized with opposite dynamics of macroeconomic indicators, influence of these factors when these crises are pooled in one regression will be cancelled out and indicator will seem insignificant. Thus wrong model specification is responsible for misleading results.
In construction of a more appropriate model we make a scenario hypothesis. For the purposes of the present research it is important to note the fact that macroeconomic literature reports important links between monetary stabilization, credit expansion and devaluation. Easterly (1996), Khamis (1996) and Stone (1998) showed that successful stabilizations are followed by rapid increase of money in real terms (or, otherwise stated, credit expansion). In the longer term there is particularly high rate of devaluations among Exchange-rate-based-stabilizations - ERBS (Hamann, 1999). And for devaluations preceding credit growth is also a contributing factor. In complete version it becomes a sort of high inflation - stabilization - devaluation - high inflation sequence (see e.g. Dornbusch, Sturzenegger, and Wolf (1990)).
Most risky for banking system stages are:
(1) inflation stabilization - typically first one or two years since stabilization program started;
(2) end of post-stabilization credit expansion - 3rd through 5th years since start of stabilization program
(3) devaluation - it is likely to follow ERBS. Timing is very uncertain, but typically devaluation is preceded by slow down of credit expansion. It may happen even 15 years after start of stabilization, but most devaluations happen in 10 years period.
We selected 3 key macroeconomic risk indicators - inflation rate change, credit growth and real exchange rate change. During monetary stabilization dynamics of these three indicators tends to be connected and all 3 affect banking system soundness in its own way.
During inflation stabilization inflation rate slowdown leads to fall in interest rates and decrease of interest rates spread. As a result reduction of real net interest income along with other inflationary sources of banks' income may lead to losses and crisis if banks' falling real income does not cover constant operating expenses. In this conditions credit growth increases real income base of banks and thus decreases probability of crisis.
On the next stage increase in loan portfolio fuelled by monetary stabilization and recovery of real sector leads to accumulation of credit risks by banks. If at the same time stabilization is achieved by fixing exchange rate competitiveness of real sector soon deteriorates and loans accumulated with continued growth expectations may sour and lead to a banking crisis. Thus we expect conjunction of credit growth and real exchange rate appreciation to be most dangerous.
Finally we insist that post-devaluation banking crisis is a special case. Massive devaluation almost by definition leads to a banking crisis. It may be triggered by external borrowings of financial sector (that lead to low or even negative net foreign assets of the banking system), by failure of enterprises working on internal market to service their foreign-exchange loans and other factors. Devaluation is typically preceded with credit growth slow down of even credit contraction, thus in contrast with the previous type of banking crisis here decrease of real exchange rate is accompanies with low or modes credit growth.
Data set available allows us to test this model on data for cross-country sample.
For cross-country sample we expect significant differences between coefficients before risk variables in the model for different crises types.
4. Theoretical model
Selection of important macroeconomic factors and their influence on banking soundness should start from microeconomic analysis of bank operations. Most studies use more of less complicated bank profitability model (see Shaffer, 1993; Dmitriev et.
al., 1996). Its most general form may be written as follows:
P Fr L i C
iC× − L× − ×
= Π
∑
∑ ∑
∑
+−
×
−
× +
× +
−
×
−
× +
= ×
k j
k
$, k k
$C, k
$, j
p, j j
RC, j
R,
C C C
1) J ) d 1 ( ) r 1 ((
C 1) J ) d 1 ( ) r 1 ((
i C
∑
∑ ∑
∑
+
−
× +
× +
−
× +
= ×
k j
k
$, k
$L, k
$, j
p, j RL, j
R,
L L L
1) J ) r 1 ((
L 1) J ) r 1 ((
i L
П - bank profit
C - loans of banks (in rubles - R, in foreign currency - $) rRC,j - real interest rate on ruble loans (j-th type)
r$C,k - interest rate on dollar-denominated loans (k-th type)
d - default rate for loans in rubles (j) and foreign currency (k) L - liabilities of banks (in rubles - R, in foreign currency - $) rRL,j - real interest rate on ruble liabilities (j-th type)
r$L,k - interest rate on dollar-denominated liabilities (k-th type) Jp,j - inflation rate for the period
J$,k - rate of currency depreciation for the period Fr - net non-interest income (expense) in real terms P - price level (accumulated inflation)
Model used in this study is basically the same and is recomposed to identify influence of particular factors.
Influence of Inflation rate change
The model could be recomposed to make explicit effect of inflation rate change.
P Fr P Lr 1) - Jp ) i 1 ((
P Cr 1) - Jp ) i 1
(( + Cr × × × − + Lr × × × − ×
= Π
П - bank profit
C - loans of banks in real terms L - liabilities of banks in real terms
Fr - net non-interest income (expense) in real terms rCr - real interest rate
rLr - real interest rate liabilities Jp,j - inflation rate for the period P - price level (accumulated inflation)
This simpler model shows that all the rest equal higher rate of inflation leads to higher bank profits. Inflationary environment typically leads to decrease of real assets and real interest rates, but operating expenses tend to grow in teal terms. As a result of several years of high inflation bank's operating expenses remain the same if not grow,
but banks real assets decrease. When inflation rate falls banks are unable reach pre- inflation equilibrium and are likely to suffer losses. Banking crisis in these conditions is likely to happen due to loss of capital when inflation goes down. The other factor that undermines banks' profitability is decrease of inflation-driven sources of income - income from currency trading and float on payments.
Influence of Real exchange rate change
Influence of real exchange rate change on banking soundness comes from two effects: net foreign currency position income (net foreign currency revaluation) and influence of real exchange rate change on credit quality and resulting loan loss performance.
The general identity above may be rewritten as sum of net foreign exchange income, net interest income less reserves for loan losses and operating expenses:
F RLL NII
NFEI+ − −
= Π
П - bank profit
NFEI - net foreign exchange rate income NII - net interest income
RLL - reserves for loan losses
F - net non-interest income (expense)
Two of the parts of the rewritten equation (namely NFEI and RLL) depend on real exchange rate change.
Real exchange rate appreciation negatively affects banks' soundness through credit quality. Appreciation decreases profitability of national producers and undermines quality of banks' loan portfolio. In fact appreciation is often accompanied with high credit growth financed by external borrowings which leads to accumulation of foreign debts of government, banks and corporations. This trend sets the stage for
possible devaluation. These borrowings in many cases lead to negative value of net foreign exchange assets.
The other component influenced by exchange rate change is net foreign currency income. It is functionally dependent on exchange rate change and net foreign currency assets. Thus both variables are explicitly included in our model. More dangerous are cases of devaluation, which are usually preceded by slowdown of credit growth.
$
$
$
$ L) J NFA J
C (
NFEI= − × = ×
NFEI - Net foreign exchange rate income C$ - loans in foreign currency,
L$ - liabilities in foreign currency, J$ - rate of real exchange rate change.
As both appreciation and depreciation are potentially dangerous for banking system soundness we have to separate them in the model. We have chosen credit growth factor for real exchange change variable splitting as two syndromes (devaluation and appreciation) are accompanied with opposite dynamics of real credit growth.
Influence of Credit growth
Credit growth in real terms also has two-fold action. From one side, during high inflation as we noted (see influence of inflation rate change) declining real assets are dangerous as they undermine real income base to cover operating expenses. So growth or even just modest decline of real loans helps to prevent banking crisis in conditions of highly inflationary environment.
On the other hand, fast credit growth is one of the most widely cited reasons for bad loans problem in many banking systems (see e.g. Gavin and Hausmann (1996), Kaminsky and Reinhart (1996), Honohan (1997)). The reason for that is quite straightforward: credit booms typically involve loan extension to new borrowers and
even sectors of the economy with no track record that could help to estimate future credit quality.
Here we also face a situation where the same dynamics of real loan portfolio have the opposite effect on bank's soundness depending on macroeconomic conditions. For credit growth variable we also introduce variable splitting. One variable will indicate credit growth in highly inflationary economy, and we expect it to decrease probability of banking crisis. The other will indicate credit growth in low-inflation economy, and we expect it to increase probability of the crisis.
On theoretical grounds we have to include in the model three macro variables:
inflation rate change, real exchange rate change and real credit growth, but the latter two variables are expected to have different effect in different circumstances. We introduce in the model variable splitting to take this into account and expect opposite signs before pair variables.
5. Data description and choice of explanatory variables
Sample description
Cross-country analysis is done on the basis of data for high inflation economies.
"Inflation stabilization" following approach of Easterly (1996) was defined as a movement from an "inflation crises" to "non crises" period where the former is defined as a period of at least two consecutive years with inflation above 40% percent and the latter as a period of at least two consecutive years with inflation below 40% percent.
Easterly used a 40% threshold, as 40% threshold level was found by Bruno and Easterly (1995) to be useful in discriminating between periods of very high inflation and moderate to low inflation. Two year minimum is used to eliminate spikes of inflation due purely to one-time price shocks such as changes in key import prices, devaluations
or price liberalizations. The year of stabilization ("year 0") was defined as a year when 40% threshold was crossed.
As a result using macroeconomic data from World Development Indicators CD- ROM, World Bank, for the period of 1970-1999 we got a sample of 46 countries.
Table 1. Groups of high inflation countries included in the sample Country group Country coverage
Transition economies Albania, Armenia, Azerbaijan, Belarus, Bulgaria, Croatia, Estonia, Georgia, Kazakhstan, Latvia, Lithuania, Macedonia, Poland, Romania, Russia, Ukraine
Latin America Argentina, Bolivia, Brazil, Chile, Dominican Republic, Ecuador, Jamaica, Mexico, Nicaragua, Peru, Uruguay, Venezuela
Africa Angola, Congo Democratic Republic, Ghana, Guinea-Bissau, Mozambique, Nigeria, Sierra Leone, Uganda, Zambia
Other Iceland, Indonesia, Israel, Lao People's Democratic Republic, Mongolia, Turkey, Yemen Arab Republic
Dating of banking crises is quite complicated. To start, researchers have to identify situations that can be termed full-fledged banking crises. Stresses in the banking system are very difficult to quantify. The data necessary for making an as- sessment are generally not available and, as a result, dating of banking crises must rely on events such as the closure of banks and official support for (and/or government takeover of) financial institutions. Generally, banking sector weaknesses emerge because of deterioration in asset quality. Reliable and timely data on nonperforming assets is not always available and even indirect evaluations of asset quality require in- formation on bankruptcies, exposures of financial intermediaries to different sectors, and movements in real estate and other asset prices—information that is generally not available in many developing and transition economies.
The other intrinsic problem with data on banking crises is linked to arbitrary nature of dating the crises. Country data on banking crises are based on case studies that usually do not use common methodology and report uniform data to make cross-country comparisons possible. As a result for banking crisis detection one needs to rely not on quantitative, but qualitative data collected from numerous sources, mostly case studies of banking crisis. In mid 1990s there were undertaken several quite extensive attempts to collect such data in compact form. These studies now serve as a primary source of data on banking crisis.
Dating of banking crises is based on two such studies: Lindgren, Garcia and Saal (1996) and Caprio and Klingebiel (1999), which is an updated version of original data set for 1996 study. These studies employ different definitions of banking crisis, but both discern two types of banking distress: significant problems and banking crisis.
Demirguc-Kunt A. and Detragiache E. (1997) proposed a set of formal criteria to diagnose a crisis:
1. The ratio of non-performing assets to total assets in the banking system exceeded 10%;
2. The cost of rescue operation was at least 2 percent of GDP
3. Banking sector problems resulted in a large scale nationalization of banks;
4. Extensive bank runs took place or emergency measures such as deposit freezes, prolonged bank holidays, or generalized deposit guarantees were enacted by the government in response to the crisis.
A bit different approach is used by Lindgren, Garcia and Saal (1996). They distinguish two type of banking problems "crisis" and "significant". The former is diagnosed in "cases where were runs or other substantial portfolio shifts, collapses of financial firms, or massive government intervention took place", the latter is defined as
"extensive unsoundness short of a crises". The authors acknowledge some degree of judgment in this classification, but considering fragmented data from case studies they assets this remains the only option available for off-site researcher.
Caprio and Klingebiel (CK) collected data on 165 episodes of bank insolvency sincethe late 1970s, of which 114 are episodes of systemic banking crisis (much or all of bank capital were exhausted) in 93 countries and 51 episodes of borderline and non- systemic banking crisis in 46 countries. "It relies upon the assessment of a variety of finance professionals. Only published sources or interviews with experts familiar with individual episodes were employed. The dates attached to the crises reviewed those generally accepted by finance experts familiar with the countries. Some judgment has gone into the list [of systemic episodes]."
Lindgren, Garcia, and Saal (LGS) collected data of IMF desk economists, Sheng (1996), Caprio and Klingebiel (1996), Sundararajan and Balino (1991), and various official and news publications. Two general classes of bank distress are identified:
"crisis" and "significant" problems. The authors acknowledge some degree of judgment in these classifications, but in general "refer to cases where there were runs or other substantial portfolio shifts, collapses of financial firms, or massive government intervention, as crises. Extensive unsoundness short of a crisis is termed significant."
Both studies generally specify starting and ending dates at annual frequency, although in several cases crises that have emerged recently are treated as ongoing. Thus most of the crises have definite length, for some there is also a figure of cost in percent of GDP. Caprio and Klingebiel (1999) present it in more uniform and complete way.
To our opinion data of Caprio and Klingebiel (1999) is better then that of Lindgren, Garcia and Saal (1996) not only because it is closer to our days but also their definition of crisis dates is more concrete. Lindgren, Garcia and Saal (1996) tend to
overextend crises. For example, in Russia crisis is dated as "1992-present". Obviously this is too rough definition, as most on-site researches report a banking boom in 1992- 1994 and banking crisis in 1995 with fast recovery in 1996. Caprio and Klingebiel in contrast report two crisis years: 1995 and 1998. For these reasons data on crisis episodes in this study comes mostly from Caprio and Klingebiel (1999) with reference to Lindgren, Garcia and Saal (1996) in case of doubts.
Table 2. Differences in dating crises between Caprio and Klingebiel (1999) and Lindgren, Garcia and Saal (1996)
Number of crises 1. No differences in starting date and length 83
2. The same starting date but different length 6 3. Different starting date of the crisis 38 4. Crises that appear only in LGS (1996) sample 27 5. Crises that appear only in CK (1999) sample 32 Total number of crises 186
Banking crisis variable
Data from these two studies allowed to select and date banking crises in our sample. All years classified as "crisis" in Lindgren, Garcia and Saal (1996) and/or Caprio and Klingebiel (1999) except starting year were omitted from regression, starting year was marked as crisis and 1 value and 0 otherwise.
Table 3. Sample of banking crises.
Country Period of crisis Type Length, Cost, start end * years % of GDP Albania 1992 present 1
Angola 1991 present 1
Argentina 1980 1982 2 3 4.0
Argentina 1989 1990 2 2
Argentina 1995 2 1
Country Period of crisis Type Length, Cost, start end * years % of GDP
Armenia 1994 present 1
Azerbaijan 1995 present 1
Belarus 1995 present 1
Bolivia 1986 1987 1 2
Bolivia 1994 present 1 4.2
Brazil 1994 present 1
Bulgaria 1991 present 2
Chile 1981 1987 2 7 29.0
Congo, DR 1994 present 2
Croatia 1995 1 1
Dominican Republic 1992 present 1
Ecuador 1995 present 1
Estonia 1992 1995 2 3 1.8
Georgia 1991 present 1
Ghana 1983 1989 1 7 3.0
Guinea-Bissau 1996 1
Guinea-Bissau 1988 1990 1
Iceland 1985 1986 1 2
Iceland 1993 1 1
Indonesia 1992 present 1 2.0
Israel 1983 1984 1 2
Jamaica 1994 present 1
Kazakhstan 1991 1995 1 4 4.5
Lao PDR early 1990s 1
Latvia 1995 present 2
Lithuania 1995 present 2
Macedonia 1993 1994 2 2
Mexico 1982 2 1
Mexico 1994 present 2 6.5
Mongolia 1991 present 1
Mozambique 1988 1993 1 6
Country Period of crisis Type Length, Cost, start end * years % of GDP
Mozambique 1994 1995 1 2
Nicaragua late 1980s present 1
Nigeria 1991 1995 1 5
Peru 1983 1990 1 8
Poland 1991 present 1
Romania 1990 present 1
Russia 1995 1
Russia 1998 present 1
Sierra Leone 1990 1 1
Turkey 1982 2 1
Turkey 1991 2 1
Turkey 1994 1 1 1.0
Uganda 1990 present 1
Ukraine 1994 present 1
Uruguay 1981 1985 2 5
Venezuela 1994 present 2 17.0
Yemen Arab Republic 1996 1
Zambia 1994 present 1
* - type of crisis: 1 - significant, 2 - crisis Sources:
1. Demirguc–Kunt A., Detragiache E. The Determinants of Banking Crises in Developing and Developed Countries. //IMF Staff Papers, Vol.45, N1, March 1998b.
2. Hardy D.C., Pazarbasioglu C. Leading Indicators of Banking Crises: Was Asia Different? //IMF working paper, June 1998
3. Frydl Ed.J. The Length and Cost of Banking Crises //IMF Working Paper WP/99/30. - 1999.
4. Lindgren C.-J., Garcia G., Saal M. I. Bank Soundness and Macroeconomic Policy, Washington: IMF, 1996;
Explanatory variables
One of the most difficult tasks was to design variables that will capture in one figure several aspects of macroeconomic changes.
Inflation rate change variable should reflect both reduction of inflation rate in the current year relative to several previous ones. Ratio of inflation rate in year t to inflation rate year t-1 does not differentiate between shift from low to very low inflation and very high and high inflation. Of the several variables tested the best results were obtained from one with this definition: Change of inflation rate in 3 years (infl3) – ratio of inflation rate (π) in the current year to average inflation rate index for the period of 3 previous years:
3 t-1 t-2 t-3
t
) 1
( ) 1
( ) 1
3 (
π π
π
π
+
× +
×
= + Infl
Devaluation variable was more straight forward: it was defined as Real Exchange Rate change – ratio of exchange rate index (1+et) in the current year to inflation rate index (1+πt):
) 1 (
) 1 J (
t t RER
= + + π e
Credit growth variable was calculated as difference of Credit (Cr) to GDP ratio in the current year and the same ratio 2 years prior. 2 year period has proved to perform best in the model.
2 t
2 t t
t
−
−
−= GDP
Cr GDP
CrGr Cr
Based on economic reasoning we expect different sign before credit growth and Real Exchange Rate change index variables in different conditions. To capture this effect we have split these two variables into 2 ones.
Credit growth effect should decrease probability of crisis (or at least not increase it) during high inflation stage. As we selected as a threshold for high inflation 40% p.a.
rate we used it also for splitting variables.
(1) Credit growth in high inflation environment:
<=
= >
. .
% 40 0
. .
% 40 1 ,
a p CPO
a p CPI
CrGr CrGr
(2) Credit growth in low inflation environment:
>
<=
= 0 40 % . .
. .
% 40 2 ,
a p CPO
a p CPI
CrGr CrGr
Similarly, we expect different sign before Real Exchange Rate change variable when credit growth flattens and when it accelerates. Real Exchange Rate appreciation tends to decrease credit quality (thus, increasing probability of crisis), when combined with credit growth it is an early warning for future crisis. Real Exchange Rate depreciation which is equally risky is usually preceded by credit growth slowdown and sometimes even decrease in lending. . This two opposite effects tend to offset each other if real exchange rate change is included in regression as one variable. We split Real exchange rate change variable variable into two ones based on the following rule:
(1) Real Exchange rate change accompanied by high Credit growth:
<=
= >
A CrGR
A CrGr RERchg
REchg
0 1 ,
(2) Real Exchange rate change accompanied with modest Credit growth or credit contraction:
=>
= <
A CrGR
A CrGr RERchg
REchg
0 2 ,
A – parameter reflecting threshold value for high credit growth, our tests proved 5% of GDP change in 2 years to be the best parameter’s value.
Other variables were included in regression to account for other influencing factors that may on their own affect bankintg system vulnerability to crisis.
Liquid assets of banks to assets ratio – to account for the effect of initial banking system’s liquidity.
We tried a proxi of adequacy ratio - capital to assets ratio (from IMF grouping of accounts of credit organizations), but regression showed no connection. The probable explanation is methodological fault of calculation that has little to do with IAS or GAAP. E.g. in Russia after 1998 crisis this indicators only slightly dropped, in fact it dropped much faster in relatively safe 1995-1997.
Net foreign assets variable of banking system and monetary authorities as % of GDP to measure banking system’s and/or monetary authorities ability to counter adverse trends in forex market.
Liquid liabilities (M3) as a % of GDP is supposed to measure effect of depth of financial system. In theory effect of this variable is ambiguous as low level of financial intermediation may make banking systems too vulnerable to minor changes in macroeconomic environment. On the other hand it may diminish influence of banking system in the economy to serving overly profitable sectors or external support readily available as it would require modest amount of funds.
Summary of the macroeconomic variables used is in the table 4.
Table 4. Variables used in cross-country estimation Variable
name
Description Comment Cri Crisis dummy
(dependent variable)
1 – crisis, 0 – non-crisis
Infl3 Stabilization variable CPI in year t / average CPI in years t-1 to t-4 CrGr1 Credit growth in high
inflation environment
Credit to GDP ratio in year t less Credit to GDP ratio in year t-4, if CPI is over 40% p.a., and 0 othervise
CrGr2 Credit growth in low Credit to GDP ratio in year t less Credit to GDP
Variable name
Description Comment inflation environment ratio in year t-4, if CPI is below 40% p.a., and
0 othervise RERchg1 Real echange rate
change accompanied with fast credit growth
Exchange rate index / Inflation rate index, if value of CrGr variable is over 5%, and 0 othervise
RERchg2 Real echange rate change accompanied with modest credit growth
Exchange rate index / Inflation rate index, if value of CrGr variable is below 5%, and 0 othervise
LA_A Liquid assets of banks to assets ratio
Source: World Development Indicators NFA_GDP Net foreign assets
variable of banking system and monetary authorities as % of GDP
Source: World Development Indicators
M3_GDP Liquid liabilities (M3) as a % of GDP
Source: World Development Indicators
6. Model specification
In our model for cross-country estimation probability of banking crisis is explained by a vector of risk factors X.
) ( )
( Crisis f X
P =
Most suitable for this type economic model is simple logit model where dependent variable denotes crisis. This type of model was used in most studies to discern between crisis and non-crisis period. But here in contrast we split some variables to get separate variables for different crisis types.
Econometric model formulation is
j
crisis
iP ( ) = α x + ε
P – probability of banking crisis, X – vector of explanatory variables,
In logistic functional form the model was specified:
∑
==
nj i
crisis
iP
0
exp( ) ) ) exp(
( x
x α α
The model is estimated by maximum likelihood.
7. Estimation results
Estimation results are presented in Appendix 1. Regressions largely support presented hypotheses.
At first stage we tried to mimic results of the previous studies by running conditional fixed-effects logit regression on the sample without variable splitting. In this calculation the only highly significant variable is inflation rate change, while credit growth is only marginally significant. Real Exchange Rate change is as in most other studies insignificant. This result is in line with one of the cited authors. But then we estimated on the same data the model presented in this study and obtained completely different results.
As before inflation rate change remained significant at 95% confidence level. This is a result usually obtained in other studies. After splitting both Credit Growth variables (CrGr1 and CrGr2) increased significance but received opposite signs.
Credit growth in highly inflationary environment (CrGr1) has negative sign. This is in line with our observation that credit growth supports income base in real terms, or in case of highly inflationary environment - credit contraction decreases it thus making crisis more profitable. On the other side credit growth in low inflation environment
(CrGr2) is also a clear risk factor. Relatively low significance may reflect the fact that part of the total influence is attributable to the action of more important factor - real exchange rate appreciation (RERchg1).
One of the searched results is significance of both split Real Exchange Rate change variables - RERchg1 and RERchg2 in contrast with insignificance of the single variable in the previous regression. When we isolated episodes of low credit growth (RERchg2) from episodes of high credit growth (RERchg1) we have in effect isolated devaluation from appreciation. As a result devaluation variable (RERchg2) obtained high significance with expected positive sigh (faster growth of foreign exchange rate then inflation rate in conditions of low credit growth does lead to crisis). A pair variable - Real Exchange Rate Change during high credit growth (RERchg1) also became significant with opposite negative sign. Again it also runs in line with our previous theoretical discussion that credit growth accompanied with real exchange rate growth is a warning factor.
Our variables that control for other factors of banking system risk in both regressions showed expected signs but with different significance. Excess liquidity does decrease probability of crisis. In fact even banking system insolvency is unlikely to erupt in a banking crisis if there are no liquidity problems. Some of post-inflationary banking crises are likely to be missed if banking system inherited from high inflation times fat liquidity cushion.
Sign before depth of financial system variable measured by M3 to GDP ratio (M3_GDP) that on theoretical grounds might have appeared with both signs turned negative. In shows that smaller financial systems are more vulnerable. Possibly it reflects also the effect of maturity of financial system which often implies less variation of the main financial indicators. Although this variable is significant only at 90% level.
The 3rd control variable - Net foreign assets of financial system to GDP ratio (NFA_GDP) although has shown expected sign (negative) is not very significant.
Possibly, it is a result of the fact that this factor is important only for one of the 3 selected types of the crisis - namely post-devaluation crisis - and has little to do with the rest two.
In conjunction with this calculation with split variables, results of traditional models may be interpreted very differently. Usually it is said that influence of devaluation is captured by other "structural" variables. In our model coefficients before real exchange rate change variable took opposite signs depending on dynamics of other variables, i.e. the type of the crisis. In simpler model influence of devaluation variable may be just netted-out as it works differently depending on circumstances.
8. Conclusions
The results of the present research confirm the scenario hypothesis of banking crisis development. In contrast with traditional econometric modeling of banking crises in this study we suggested that some factors of banking crisis act not independently but in conjunction with other factors. Moreover in some scenarios influence of the given factor of risk may be opposite when seen in combination with other factors.
Other econometric models that do not distinguish between crisis scenarios miss very important sides of the problem and thus obtain sometimes disappointing results.
Macroeconomic factors work in combination to produce banking crisis and not independently as assumed in other models. And some of the risk factors take opposite signs for different types of the crisis. Models that do not recognize different crisis types tend to select as significant risk factors only those those act always in the same manner.
Most notorious example of missing evidence is lack of support for devaluation as a risk factor of the crisis. That happens because real exchange rate appreciation is a risk factor
in combination with fast credit growth. On the other hand sharp real exchange rate depreciation that is typically preceded with credit growth lowdown or even contraction triggers banking crisis as shows the model presented in this paper. Thus, opposite dynamics of real exchange rate in the run up to the crisis both act as risk factors but in different crisis syndromes but are "cancelled out" if pooled in one regression.
The other inference from the model is that authorities of high inflation countries must be prepared to fight not one as it is usually suggested but three different crises that may occur all in sequence or some of potentially dangerous developments may not end in full-fledged banking crisis. The possible reason for missing some types of the crisis may be other factors that may soften negative impact. For instance, high liquidity may well help bank to suffer losses without affecting their ability to serve deposits in the short-run.
The full sequence of crises is:
- post-inflation banking crisis that happens just after start of stabilization,
- post-stabilization crisis in the end of the period of fast credit growth in conditions of relative price stability, if accompanied with exchange rate appreciation credit growth is very likely to create conditions of banking crisis development,
- post-devaluation crisis that erupts if stabilization ends with devaluation.
0 1 2 3 4 5 6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Post-devaluation Post-stabilization Post-inflation
year after stabilization (0) number of crises
Types of banking crisis:
Figure 1. Distribution of banking crises incidence relative to stabilization year in high inflation-economies.
Graphic illustration of distribution of banking crises relative to stabilization date in our sample is presented on figure 1.
All post-stabilization crises are concentrated in the 2nd to 5th years of stabilization.
Devaluation crises are the last in series of possible crises. They are spread over 13-year period in our sample, which is consistent with macroeconomic literature.
Russian experience well fits this scheme. Russia had all three crises. Russian stabilization started early in 1995 with effective fixing of exchange rate with following 10% nominal ruble appreciation. Post-inflation banking crisis in Russia was manifested in August 1995 by interbank loans crisis. This crisis started process of massive license withdrawal in 1995-1996 (in more detail see Dmitriev et. al. (1996)). Post-stabilization crisis in the beginning of 1998 led to some famous banks failures even before August 1998 devaluation, but devaluation has precipitated even more spectacular crisis that is much more known.
Bank supervision and adequate bank management are essential to minimize potential losses. Regulators and bankers must know what risks stabilization brings and what they must prepare themselves to.
9. Bibliography
Cross-country research (listed by year of publication)
1. Sundararajan V., Balino T. (1991), Banking Crises: Cases and Issues, Washington: International Monetary Fund, pp. 1-57.
2. Caprio, G., Klingebiel D. (1996), "Bank Insolvencies: Cross-Country Experience," Policy //Research Working Paper 1620, Washington, D.C.: The World Bank.
3. Kaminsky G., Reinhart C. (1996) ‘The twin crises: The causes of banking and balance-of-payments problems’, //International Finance Discussion Paper No 544, Washington: Board of Governors of the Federal Reserve System).
4. Demirguc-Kunt A., Detragiache E. (1997) "The Determinants of Banking Crises:
Evidence from Developing and Developed Countries" //IMF Working Paper, WP/97/106.
5. Honohan P. (1997), "Banking System Failures in Developing and Transition Countries: Diagnosis and Prediction," Basle: BIS.
6. Eichengreen B., Rose A.K. (1998), "Staying Afloat When the Wind Shifts:
External Factors and Emerging-Market Banking Crises. //NBER Working Paper, No 6370.
7. Hardy D., Pazarbasioglu C. (1998) ‘Leading indicators of banking crises: Was Asia different?’ //IMF Working paper, 98/91.
8. Kaminsky G. (1998) "Currency and Banking Crises: The Early Warnings of Distress," //International Finance Discussion Paper No 629, Washington: Board of Governors of the Federal Reserve System.
9. Demirguc-Kunt A., Detragiache E. (1999) "Monitoring Banking System Fragility:
A Multivariate Logit Approach" //IMF Working Paper, WP/99/147.
10. Hardy D., Pazarbasioglu C. (1999) ‘Determinants and Leading indicators of banking crises: Further Evidence’ //IMF Staff Papers, Vol. 46, No 3.
Studies of individual bank failures (listed by year of publication)
11. Cole R.A., Gunther J.W. (1993) 'Separating the Likelihood and Timing of Bank Failure' //Finance and Economic Discussion Series, Federal Reserve Board, No.
93-20.
12. Cole R.A., Cornyn B.G., Gunther J.W. (1995) 'FIMS: A New Monitoring System for Banking Institutions' //Federal Reserve Bulletin, Federal Reserve Board, No.
81, pp. 1-15.
13. Gonzales-Hermosillo B., Pazarbasioglu C., Billings R. (1997) Determinants of Banking System Fragility: A Case Study of Mexico’ //IMF Staff Papers, Vol. 44, No. 3.
14. Gonzales-Hermosillo B. (1999) Determinants of Ex-Ante Banking System Distress: A Macro-Micro Empirical Exploration of Some Recent Episodes //IMF Working Paper WP/99/33.
Data on banking crisis episodes
15. Lindgren, C.-J., Garcia G., Saal M.(1996) Bank Soundness and Macroeconomic Policy. Washington: IMF.
16. Caprio, G., Klingebiel D. (1999), "Episodes of Systemic and Borderline Financial Crises", Washington: The World Bank.
Other studies
17. Akerlof, G., P. Romer (1993) ‘Looting: The economic underworld of bankruptcy for profit’, Brookings Papers on Economic Activity, 2, pp. 1 – 73;
18. Alexander, W., J. Davis, L. Ebrill, C.-J. Lindgren (1997) Systemic Bank Restructuring and Macroeconomic Policy. Washington, D.C.: IMF.
19. Bruno M., Easterly W. Inflation Crises and Long-Run Growth. /NBER Working Paper, No 5209, 1995
20. Calomiris, Charles and Gary Gorton (1991), "The Origins of Banking Panics:
Models, Facts and Bank Regulation," in R. Glenn Hubbard (ed.). Financial Markets and Financial Crises, Chicago: University of Chicago Press, pp.109-174.
21. Claessens S. (1996) Banking Reform in Transition Countries, IABRD Conference Paper.
22. Demirguc-Kunt A., Detragiache E. (1998) Financial Liberalization and Financial Fragility - IMF Working Papers № 83.
23. Demirguc-Kunt, Asli and Enrica Detragiache (2000) "Does Deposit Insurance Increase Banking System Stability" //IMF Working Paper, WP/00/3.
24. Diamond, Douglas and Hump Dybvig (1983), "Bank runs. Liquidity and Deposit Insurance," Journal of Political Economy 9, pp.40l-419.
25. Dmitriev, Matovnikov, Mikhailov, Sycheva, Timofeev, Warner (1996) Russian Banks on the Eve of Financial Stabilization - St-Petersburg.
26. Dornbusch, Rudiger, and Alejandro Werner (1994) "Mexico: Stabilization, Reform, and No Growth," Brookings Papers on Economic Activity, No. 1, pp. 253-315.
27. Dornbusch, Rudiger, Federico Sturzenegger, and Holger Wolf (1990) "Extreme Inflation: Dynamics and Stabilization," Brookings Papers on Economic Activity, No. 2, pp.1-84.
28. Drees, B., C. Pazarbasioglu (1998) ‘The Nordic banking crises. Pitfalls in financial liberalization?’/ IMF Occasional paper, 161.
29. Easterly W. (1996) When is Stabilization Expansionary? Evidence from High Inflation / Economic Policy, vol. 22.
30. English W. (1996) Inflation and Financial Sector Size //Federal Reserve Board Finance and Economics Discussion Series No 16.
31. Frydl Ed.J. (1999) The Length and Cost of Banking Crises //IMF Working Paper WP/99/30.
32. Gavin, Michael and Ricardo Hausmann (1995), "The Roots of Banking Crises:
The Macroeconomic Context," in Ricardo Hausmann and Liliana Rojas-Suarez (eds).
Banking Crises in Latin America, Baltimore: Johns Hopkins University Press, pp.27-63.
33. Goldstein, Morris and Philip Turner (1996), "Banking Crises in Emerging Economies: Origins and Policy Options," BIS Economic Papers DO.46, Bask: Bank for International Settlements (October).
34. Hamann A.J. (1999) Exchange-Rate-Based Stabilization: A Critical Look at the Stylized Facts. / IMF Working Paper WP/99/132..
35. Heffernan, S. (1996) Modern Banking in Theory and Practice. John Wiley &
Sons.
36. International Monetary Fund (1998) World Economic and Financial Surveys International Capital Markets: Developments, Prospects and Key Policy Issues, Washington D.C.: IMF.
37. Juan de, A. (1995), "The Roots of Banking Crises: Microeconomic Issues and Issues of Supervision and Regulation," paper presented to the IADB/G-30 Conference on Banking Crises in Latin America.
38. Khamis M. (1996) Credit and Exchange Rate-Based Stabilization /IMF Working Paper WP/96/51.
39. Matovnikov (2000) Functioning of Russian Banking System in Conditions of Macroeconomic Instability - Moscow: IET.
40. Mishkin, Frederic S. (1996), "Understanding Financial Crises: A Developing Country Perspective," NBER Working Paper no. 5600 (May).
41. Pazarbasioglu, C. (1997) ‘A credit crunch? Finland in the aftermath of the banking crises’, IMF Staff Papers, 44, pp. 315 – 327;
42. Rojas-Suarez, Liliana and Steven R. Weisbrod (1994), "Financial Market Fragilities in Latin America: From Banking Crisis Resolution to Current Policy Challenges," IMF Working Paper WP/94/117.
43. Santanella J.A., Vela A.E. (1996) The 1987 Mexican Disinflation Program; An Exchange Rate-Based Stabilization? /IMF Working Paper WP/96/24.
44. Stone, M. (1998) Financial Infusion and Exiting from a Money Rule, IMF Working Paper WP/98/31.
45. Sundararajan, V., T. Balino (1991) Banking Crises: Cases and Issues.
Washington, D.C.: IMF.
46. Vegh, Carlos (1995) "Real Effects of Exchange Rate-Based Stabilizations: An Analysis of Competing Theories" International Monetary Fund.
Appendix 1. Estimation results for cross-country sample The model: conditional fixed-effects logit regression
Number of obs = 814 Number of groups = 46 Obs per group: min = 9
avg = 17.7
max = 29
LR chi2(3) = 4.85
Log likelihood = -114.43704 Prob. > chi2 = 0.0042 Cri Coef. Std. Err. z P>|z| [95% Conf. Interval]
Infl3 -0.0249488 0.012481 -1.999 0.046 -0.04941 -0.00049 CrGr1 -2.695608 1.310504 -2.057 0.040 -5.26415 -0.12707 CrGr2 0.0405295 0.022762 1.781 0.075 -0.00408 0.085141 RERchg1 -3.299308 1.005192 -3.282 0.001 -5.26945 -1.32917 RERchg2 3.263575 1.19656 2.727 0.006 0.918361 5.608789
LA_A -.0375454 .0159546 -2.353 0.019 -.0688158 -.0062751 NFA_GDP -0.34024 0.236622 -1.438 0.150 -0.80401 0.123527
M3_GDP -0.0399604 0.023413 -1.707 0.088 -0.08585 0.005928 Alternative model (no variable splitting): conditional fixed-effects logit regression
Number of obs = 814 Number of groups = 46 Obs per group: min = 9
avg = 17.7
max = 29
LR chi2(3) = 7.54
Log likelihood = -103.74085 Prob. > chi2 = 0.0286 Cri Coef. Std. Err. z P>|z| [95% Conf. Interval]
Infl3 -0.02687 0.012831 -2.094 0.036 -0.05202 -0.00172 CrGr 0.0215886 0.014579 1.481 0.139 -0.00699 0.050163 RERchg -0.0059203 0.027164 -0.218 0.827 -0.05916 0.047321
LA_A -0.075672 0.033434 -2.263 0.024 -0.141201 -0.010143 NFA_GDP -0.02668 0.01865 -1.431 0.153 -0.00987 0.063233
M3_GDP -1.174255 0.649407 -1.808 0.071 -2.44707 0.098559
Appendix 2. Banking crises in 1970-1999
Country Period of crisis Type Length, Cost, start end * years % of GDP Albania 1992 present 1
Algeria 1990 1992 1 3
Angola 1991 present 1
Argentina 1980 1982 2 3 4.0
Argentina 1989 1990 2 2
Argentina 1995 2 1
Armenia 1994 present 1
Australia 1989 1992 1 4
Azerbaijan 1995 present 1
Bangladesh 1980 present 1 4.5
Belarus 1995 present 1
Benin 1988 2 1
Buhtan 1990 present 1
Bolivia 1986 1987 1 2
Bolivia 1994 present 1 4.2
Bosnia-Herzegovina 1992 present 1
Botswana 1994 1995 1 2
Brazil 1994 present 1
Brunei Darussalam mid 1980s 1
Bulgaria 1991 present 2
Burkina Faso 1988 1994 1 7
Burundi 1994 present 1
Cambodia 1 1
Cameroon 1989 1993 2 5
Cameroon 1995 present 2
Canada 1983 1985 1 3
Cape Verde 1993 present 1
C.A.R. 1986 1992 2 7
C.A.R. 1995 present 1
Chad 1979 1983 2 5
Chad 1992 1 1
Chile 1981 1987 2 7 29.0
China 1980s present 1
Colombia 1982 1985 1 4
Congo, Democratic Republic 1994 present 2 Costa Rica 1994 present 2
Cote d'Ivoire 1988 1990 1 3
Croatia 1995 1 1
Czech Republic 1991 present 1 12.0
Denmark 1987 1992 1 6
Djibuti 1991 1993 1 3
Dominican Republic 1992 present 1
Ecuador 1995 present 1
Egypt 1991 1995 1 5
El Salvador 1989 1 1
Country Period of crisis Type Length, Cost, start end * years % of GDP Equatorial Guinea 1983 1985 2 3
Equatorial Guinea 1995 1 1
Eritrea 1994 1 1
Estonia 1992 1995 2 3 1.8
Ethiopia 1994 1995 1 2
Fiji 1995 present 1
Finland 1991 1994 2 4 8.4
France 1991 1995 1 5 0.6
Gabon 1995 present 1
Gambia, The 1985 1992 1 8
Georgia 1991 present 1
Germany 1990 1993 1 4
Ghana 1983 1989 1 7 3.0
Greece 1991 1995 1 5
Guinea 1980 1985 2 6
Guinea-Bissau 1996 1
Guinea-Bissau 1988 present 1
Gyana 1993 1995 1 3
Haiti 1991 present 1
Hungary 1987 present 1 9.0
Iceland 1985 1986 1 2
Iceland 1993 1 1
India 1991 present 1
Indonesia 1992 present 1 2.0
Ireland 1985 1 1
Israel 1983 1984 1 2
Italy 1990 1995 1 6
Jamaica 1994 present 1
Japan 1992 present 1
Jordan 1989 1990 2 2
Kazakhstan 1991 1995 1 4 4.5
Kenya 1993 1 1
Korea mid 1980s 1
Kuwait mid 1980s 2
Kuwait 1990 1991 1 2
Kyrgyz Republic 1996 1 Lao People's Democratic Republic early 1990s 1
Latvia 1995 present 2
Lebanon 1988 1990 2 3
Lesoto 1988 present 1
Liberia 1991 1995 2 5
Lithuania 1995 present 2
Macedonia 1993 1994 2 2
Madagascar 1988 1 1
Madagascar 1991 1995 1 5
Malaysia 1985 1988 2 4 4.1
Mali 1987 1989 1 3
Mali 1995 1 1
Country Period of crisis Type Length, Cost, start end * years % of GDP
Mauritania 1991 1993 1 3
Mauritius 1996
Mexico 1982 2 1
Mexico 1994 present 2 6.5
Moldova 1994 present 1
Mongolia 1991 present 1
Mozambique 1988 1993 1 6
Mozambique 1994 1995 1 2
Myanmar 1996 1
Nepal late 1980s 1
New Zealand 1989 1990 1 2
Nicaragua late 1980s present 1
Niger 1983 present 2
Nigeria 1991 1995 1 5
Norway 1987 1993 2 7 3.3
Pakistan 1980 present 1
Panama 1988 1989 2 2
Papua New Guanea 1989 present 1
Paraguay 1995 present 1
Peru 1983 1990 1 8
Philippines 1981 1987 2 7 13.2
Poland 1991 present 1
Romania 1990 present 1
Russia 1992 present 1
Rwanda 1991 present 1
Sao Tome and Principe 1980 present 2
Senegal 1983 1988 2 6
Sierra Leone 1990 1 1
Singapore 1982 1 1
Slovak Republic 1991 1995 1 5
Slovenia 1992 1994 1 3
Somalia 1990 2
South Africa 1985 2 1 South Africa 1989 present 1
Spain 1977 1985 2 9 5.6
Sri Lanka early 1990s 1 St. Vincent and the Grenadines 1994 present 1
Swaziland 1995 1 1
Sweden 1990 1993 2 4 4.0
Tajikistan 1996 1
Tanzania 1988 present 2 6.5
Thailand 1983 1987 2 5
Togo 1989 1991 1 3
Trinidad and Tobago 1982 1993 1 12
Tunisia 1991 1995 1 5
Turkey 1982 2 1
Turkey 1991 2 1
Turkey 1994 1 1 1.0
Country Period of crisis Type Length, Cost, start end * years % of GDP
Uganda 1990 present 1
Ukraine 1994 present 1
US 1980 1992 1 13 2.4
Uruguay 1981 1985 2 5
Uzbekistan 1993 present 1
Venezuela 1994 present 2 17.0
Vietnam 1996 1
Yemen Arab Republic 1996 1
Zaire 1991 present 1
Zambia 1994 present 1
Zimbabwe 1995 present 1
Number of observations 152 81 22
* - type of crisis: 1 - significant, 2 - crisis Sources:
1. Demirguc–Kunt A., Detragiache E. The Determinants of Banking Crises in Developing and Developed Countries. //IMF Staff Papers, Vol.45, N1, March 1998b.
2. Hardy D.C., Pazarbasioglu C. Leading Indicators of Banking Crises: Was Asia Different? //IMF working paper, June 1998
3. Frydl Ed.J. The Length and Cost of Banking Crises //IMF Working Paper WP/99/30. - 1999.
4. Lindgren C.-J., Garcia G., Saal M. I. Bank Soundness and Macroeconomic Policy, Washington: IMF, 1996;
